Fan blade on-line monitoring method and system

Abstract

The application discloses a wind turbine blade online monitoring method and system, and relates to the technical field of wind power generation operation and maintenance. The method comprises the following steps: in adjacent operation periods in a wind turbine blade operation collection process, a continuous change track of a propagation area is calculated and a candidate continuous area is determined according to a signal propagation area distribution state and a region overlapping relationship between the signal propagation areas, and propagation data is obtained; based on the candidate continuous area and the region overlapping relationship, a propagation area correlation processing is performed on a local propagation voiceprint corresponding to an operation period, and correlation data is obtained; an abnormal propagation area is determined according to a propagation correlation sequence and a region overlapping parameter; based on a propagation continuous parameter and the region overlapping parameter, a continuous area correction is performed on the abnormal propagation area, and abnormal propagation result data is output. The application solves the problem that the wind turbine blade defect positioning result jumps back and forth between different monitoring periods when the wake interference is strong, realizes stable tracking of the cross-period voiceprint migration track, and improves the time sequence coherence of defect positioning.

Description

Technical Field

[0001] This application relates to the field of wind power operation and maintenance technology, and in particular to online monitoring methods and systems for wind turbine blades. Background Technology

[0002] As the installed capacity of wind farms continues to expand, the loads on wind turbine blades during operation are becoming increasingly complex. In order to detect potential cracks or damage inside the blades in a timely manner, vibration or acoustic sensors are usually placed on the blade surface to analyze the health status of the blades by collecting signals at different time periods.

[0003] The common practice is to divide the blade surface into a fixed monitoring grid and then compare the signal characteristics collected in different cycles within each grid. If the signal in a certain monitoring grid shows a significant change, it is considered that there may be an anomaly at that location. However, this method is suitable for relatively stable wind conditions and when the blade stress does not change much. But when the wind turbine is in a densely packed wind field, the rear blades are continuously affected by the wake of the front units, making the incoming flow turbulent, and the aerodynamic load on the blades fluctuates accordingly. As a result, the spatial distribution of the signal generated by the same defect will shift when collected in different cycles, and the signal characteristics of adjacent operating cycles can easily overlap spatially. The traditional fixed grid comparison method does not consider this dynamic shift, and when matching in areas where signals overlap, it is easy to make inaccurate judgments, causing the output anomaly location to jump back and forth between different cycles, which causes confusion for maintenance personnel.

[0004] Therefore, there is a prominent problem in the existing technology: under strong wake interference, traditional methods are unable to guarantee the consistency of defect location results between different monitoring cycles, and are prone to misjudgment of location. Summary of the Invention

[0005] In view of the aforementioned problems, this application is hereby filed.

[0006] Therefore, this application provides a method and system for online monitoring of wind turbine blades, which can solve the problems mentioned in the background art.

[0007] To solve the above-mentioned technical problems, this application provides the following technical solution: Firstly, this application provides a method for online monitoring of wind turbine blades, including: During adjacent operating cycles of the wind turbine blade operation data acquisition process, the continuous change trajectory of the propagation area is calculated and candidate continuous areas are determined based on the distribution status of the signal propagation area and the regional overlap relationship between the signal propagation areas, thus obtaining propagation data. The propagation data includes propagation continuity parameters, regional overlap relationship and candidate continuous areas. Based on candidate continuous regions and region overlap relationships, propagation region association processing is performed on the local propagation voiceprints corresponding to the running cycle to obtain association data, which includes propagation association sequence and region overlap parameters. The abnormal propagation region is determined based on the propagation association sequence and the region overlap parameter; Based on the propagation continuity parameter and the region overlap parameter, continuous region correction is performed on the abnormal propagation region, and abnormal propagation result data is output.

[0008] Preferably, during the wind turbine blade operation data acquisition process, within adjacent operating cycles, based on the signal propagation area distribution and the overlapping relationship between signal propagation areas, the continuous change trajectory of the propagation area is calculated and candidate continuous areas are determined to obtain propagation data, including: Extract the position range of each signal propagation region along the blade span within adjacent operating cycles to form the signal propagation region distribution state; Based on the projected coordinates of the position range on the blade surface, the two position ranges with the closest spatial centroid distance in adjacent operating cycles are paired and associated to form a region matching pair; Calculate the coordinate difference in the spanwise direction and the projection offset in the chordwise direction of the two position ranges in the region matching pair to generate the propagation continuity parameter; Obtain the overlap ratio and boundary fit of adjacent operating cycle regions projected onto the blade surface to form the region overlap relationship; The propagation continuity parameter and the overlap ratio are periodically aligned, and regions with consistent position change directions and continuous boundary adhesion are selected and marked as candidate continuous regions.

[0009] Preferably, the step of extracting the position range of each signal propagation region along the blade spanwise within adjacent operating cycles to form the signal propagation region distribution includes: The temporal sound pressure records output by the acoustic fingerprint sensor array deployed along the blade span are obtained in each operating cycle. The data segments of each operating cycle are extracted with the blade rotation phase as the alignment reference, and the reference sound pressure record value of each acoustic fingerprint sensor measurement point is determined. The reference sound pressure recording value is assigned to the three-dimensional coordinate position corresponding to the blade surface measurement point layout diagram, and the acoustic pattern recording intensity distribution along the blade span is obtained by spatial interpolation. The recording regions within the effective signal range of the acoustic signature intensity distribution are selected, and the contour boundaries of the recording regions along the blade span are extracted and fitted to obtain the position range.

[0010] Preferably, the propagation region association processing is performed on the local propagation voiceprints corresponding to the operating cycle based on the candidate continuous regions and the region overlap relationship to obtain associated data, including: The candidate continuous regions are configured as associated nodes according to the order of the operation cycle, a periodic connection relationship is constructed, and the degree of boundary fit in the region overlap relationship is determined as the connection weight between associated nodes. Extract the frequency band distribution record and intensity decrease record of the local propagation acoustic text corresponding to the associated node, and determine the degree of feature difference between the connected associated node pairs in adjacent operating cycles based on the periodic connection relationship; The connection weights are matched with the degree of difference in the features to generate a region association weight table; Based on the aforementioned periodic connection relationship, the associated nodes connected in adjacent operating cycles are sequentially linked together to generate a node connection path; The connection weights corresponding to the node connection paths are read from the region association weight table, and the connection weights are calculated to obtain the node path score. The node path scores are sorted, and the node connection path with the highest score is selected as the propagation association sequence. The extreme values ​​of the boundary fit of the associated nodes in the node connection path are extracted as the region overlap parameters.

[0011] Preferably, the step of extracting the frequency band distribution record and intensity decrease record of the local propagation acoustic text corresponding to the associated node, and determining the degree of feature difference between connected pairs of associated nodes in adjacent operating cycles based on the periodic connection relationship, includes: The local propagation acoustic text is subjected to frequency band division processing, and the signal recording ratio in each frequency band interval is statistically analyzed to obtain the frequency band distribution record. Extract the peak sequence of the local propagation soundprint, calculate the ratio of the amplitude difference between adjacent peaks to the time interval, record the ratio as the amplitude decrease rate, and obtain the intensity decrease record; The frequency band distribution record and the intensity decrease record are combined into a feature data group, and the difference measure value between the feature data groups of related nodes in adjacent operating cycles is calculated as the feature difference degree.

[0012] Preferably, determining the abnormal propagation region based on the propagation association sequence and the region overlap parameter includes: Perform periodic signal recording superposition processing on each associated node in the propagation association sequence to obtain the cumulative value of the voiceprint feature corresponding to each associated node; Based on the extreme values ​​of boundary fitting in the region overlap parameters, an overlap interference correction coefficient is generated, and numerical conversion processing is performed on the cumulative value of voiceprint features to obtain the converted cumulative value of features. The calculated cumulative value of the feature is compared with the historical baseline sequence, and the associated nodes whose calculated values ​​are higher than the upper limit of the distribution of the historical baseline sequence within a continuous running period are selected to obtain a set of abnormal associated nodes; Calculate the coordinate difference in the blade span of adjacent associated nodes in the set of abnormal associated nodes, divide the associated nodes whose coordinate difference is less than the clustering distance threshold into spatial clustering node groups, and determine the blade span of coordinate interval covered by the spatial clustering node groups as the abnormal propagation area.

[0013] Preferably, the step of generating overlap interference correction coefficients based on the extreme values ​​of boundary fit in the region overlap parameters, and performing numerical conversion processing on the accumulated voiceprint feature values ​​to obtain converted feature accumulated values ​​includes: From the associated nodes contained in the propagation association sequence, the associated nodes that have overlapping blade surface projections within the same operating cycle are selected and determined as candidate nodes; Extract the extreme values ​​of the boundary fitting degree of the region overlap parameters corresponding to each candidate node, and use the ratio of the extreme values ​​of the boundary fitting degree to the reference fitting value as the overlap interference correction coefficient. Evaluate the continuous state of the characteristic path of the candidate node and the associated node in the adjacent running cycle, and determine the node in the candidate node whose continuous state of the characteristic path undergoes a step change as a transient interference node. The cumulative value of the acoustic signature feature corresponding to the transient interference node is removed, and the node whose feature path continues to evolve smoothly among the remaining candidate nodes is selected as the main propagation node. Based on the overlapping interference correction coefficient, the cumulative value of the voiceprint feature corresponding to the main propagation node is numerically converted to obtain the converted feature cumulative value.

[0014] Preferably, the step of performing continuous region correction on the abnormal propagation region based on the propagation continuity parameter and the region overlap parameter, and outputting abnormal propagation result data, includes: Extract the position change relationship sequence from the propagation continuity parameters, determine the expected displacement vector of adjacent operating cycles based on the position change relationship sequence, and establish a displacement evolution benchmark; Based on the displacement evolution benchmark, the sound velocity parameters of the blade composite material, and the curvature of the local surface, the original position sequence of the abnormal propagation region is subjected to position adjustment processing, and a preliminary adjustment sequence is output. The position jump point is identified by using the extreme value of the boundary fit degree in the region overlap parameter, and the stable coordinate data of the adjacent running cycles before and after the position jump point are read. Based on the stable coordinate data, continuous transition coordinates are generated. The jump coordinates in the preliminary adjustment sequence are replaced with the continuous transition coordinates. The coordinate data are arranged in the order of the running cycle to generate a position correction sequence. The position correction sequence is projected onto the blade's three-dimensional spatial coordinate system, and the anomaly propagation result data, which includes the corrected spanwise position coordinates, the anomaly extension direction, and the region confidence weight, is output.

[0015] Preferably, the original position sequence of the abnormal propagation region is adjusted by combining the displacement evolution benchmark, the sound velocity parameters of the blade composite material, and the curvature of the local surface, and a preliminary adjustment sequence is output, including: Obtain wake turbulence intensity values ​​and blade root bending moment load data from the unit's operation records; Based on the bending moment load data at the blade root, the stress distribution along the blade span is determined, and the sound velocity adjustment value corresponding to the stress distribution is determined according to the material sound velocity correspondence. The execution parameters of the position adjustment process are adjusted according to the wake turbulence intensity value, and the position adjustment process generates the preliminary adjustment sequence based on the adjusted execution parameters under the condition of wake overlap.

[0016] Secondly, this application also provides an online monitoring system for wind turbine blades, including: The trajectory construction module calculates the continuous change trajectory of the propagation area and determines candidate continuous areas based on the distribution status of the signal propagation area and the regional overlap relationship between the signal propagation areas during adjacent operating cycles in the wind turbine blade operation acquisition process, thereby obtaining propagation data. The propagation data includes propagation continuity parameters, regional overlap relationship and candidate continuous areas. The association processing module performs propagation region association processing on the local propagation voiceprints corresponding to the running cycle based on candidate continuous regions and region overlap relationships to obtain association data, which includes propagation association sequences and region overlap parameters. The anomaly detection module determines the abnormal propagation region based on the propagation association sequence and the region overlap parameter; The continuous correction module performs continuous region correction on the abnormal propagation region based on the propagation continuity parameter and the region overlap parameter, and outputs abnormal propagation result data.

[0017] Implementing this application has the following beneficial effects: This application provides an online monitoring method and system for wind turbine blades. Addressing the problem of cross-cycle misjudgment caused by dynamic spatial offset and overlap of signals when wake interference is strong, which leads to inconsistent positioning results in traditional fixed grid comparison, this application determines candidate continuous regions by calculating the continuous change trajectory of the signal propagation area in adjacent operating cycles. Based on the region overlap relationship, cross-cycle correlation processing is performed on the local propagation acoustic signature to lock abnormal propagation areas. Furthermore, continuous region correction is performed on abnormal areas by combining propagation continuity parameters and region overlap parameters. This application transforms discrete periodic grid comparison into continuous trajectory tracking and spatial correction, effectively resolving signal offset and multi-solution conflicts caused by wake overlap, ensuring strict temporal continuity of defect positioning results between adjacent monitoring cycles, and eliminating maintenance misjudgments caused by fluctuating positioning coordinates. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is an overall flowchart of the online monitoring method for wind turbine blades involved in this application; Figure 2 This is an application environment diagram of the online monitoring method for wind turbine blades involved in this application; Figure 3 This is a schematic diagram of the overall structure of the online monitoring system for wind turbine blades involved in this application; Figure 4 This is a diagram of the computer equipment used in the online monitoring method for wind turbine blades involved in this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0021] With the continuous expansion of wind farm installed capacity and increasingly dense turbine layout, monitoring the health status of wind turbine blades under complex alternating loads has become a core requirement for ensuring the safe operation of wind farms. In engineering practice, acoustic or vibration sensor arrays are typically deployed on the blade surface to continuously collect time-series monitoring data from different operating cycles. How to accurately identify the evolution trajectory of internal cracks or damage from these massive, dynamically changing monitoring signals, and achieve stable positioning across monitoring cycles, has become a key technical requirement for intelligent blade operation and maintenance and condition assessment.

[0022] Traditional blade defect monitoring methods primarily rely on dividing the blade surface into fixed monitoring grids and comparing changes in signal characteristics across different grids over different periods to determine anomaly locations. However, this method is only suitable for stable wind conditions and uniform blade stress. In densely packed wind fields, the rear blades are continuously interfered with by the wake flow of the preceding units. Turbulent airflow causes drastic fluctuations in aerodynamic loads, resulting in dynamic shifts in the spatial distribution of acoustic signature signals generated by the same defect across different periods. Signal characteristics from adjacent periods are highly likely to overlap spatially. Traditional fixed grid matching does not consider this dynamic shift and spatial overlap, easily leading to multiple conflicting solutions when comparing features in overlapping areas. This results in frequent jumps in the output anomaly location results between different monitoring periods, making it difficult to guarantee the temporal continuity of the defect trajectory and severely hindering accurate judgment and intervention decisions by maintenance personnel.

[0023] To address the limitations of existing technologies in ensuring the continuity of defect location results across cycles and the susceptibility to misjudgment under wake interference conditions, this application proposes an online monitoring method for wind turbine blades. This application calculates the continuous change trajectory of the propagation region based on the signal propagation region distribution and region overlap relationship within adjacent operating cycles to determine candidate continuous regions. Based on the candidate continuous regions and overlap relationships, cross-cycle correlation processing is performed on the local propagation acoustic signatures to generate propagation correlation sequences and region overlap parameters. The propagation correlation sequences and region overlap parameters are used to pinpoint abnormal propagation regions, and continuous region correction is performed on the abnormal regions using the propagation continuity parameters and overlap parameters. Finally, the abnormal propagation result data, including corrected spatial coordinates, abnormal extension direction, and region confidence weights, is output.

[0024] In one exemplary embodiment, such as Figure 1 As shown, an online monitoring method for wind turbine blades is provided, including: S1: During adjacent operating cycles in the wind turbine blade operation acquisition process, the continuous change trajectory of the propagation area is calculated and candidate continuous areas are determined based on the distribution status of the signal propagation area and the regional overlap relationship between the signal propagation areas, thus obtaining propagation data. The propagation data includes propagation continuity parameters, regional overlap relationship and candidate continuous areas.

[0025] In some embodiments, step S1 includes steps S11 to S15: Step S11: Extract the position range of each signal propagation area along the blade span in adjacent operating cycles to form the signal propagation area distribution status.

[0026] Step S11 includes steps S111 to S113: Step S111: Obtain the time-series sound pressure records output by the acoustic sensor array deployed along the blade spanwise in each operating cycle, extract the data segments of each operating cycle with the blade rotation phase as the alignment reference, and determine the reference sound pressure record value of each acoustic sensor measurement point.

[0027] Specifically, when the wind turbine blades rotate and cut the airflow, aerodynamic noise and defect acoustic patterns overlap in the time domain. If data from a fixed time window is directly extracted, acoustic pattern signals from the same physical location will fall into different data segments, causing cross-cycle comparison distortion. Therefore, extracting data segments based on the blade rotation phase can effectively eliminate sampling deviations caused by rotational asynchrony.

[0028] Furthermore, data segments for each operating cycle are extracted using the blade rotation phase as the alignment reference. Specifically, the moment when a specific reference point of the blade passes through the sensor array is used as the starting boundary, and sound pressure segments corresponding to a fixed duration are extracted to ensure that the reference sound pressure record value accurately reflects the structural vibration response within a single cycle, thereby providing a time-aligned data reference for spatial distribution reconstruction.

[0029] Step S112: Assign the reference sound pressure recording value to the three-dimensional coordinate position corresponding to the blade surface measurement point layout diagram, and obtain the acoustic signature recording intensity distribution along the blade span by spatial interpolation.

[0030] Furthermore, given that the sensor array can only acquire sound pressure data from dispersed measurement points and that there are physical blind spots on the blade surface, this application fills the gaps between measurement points through spatial interpolation and transforms the discrete point array into a continuous energy surface, so that the intensity distribution of the acoustic signature records presents a continuous energy fluctuation pattern to accurately characterize the radiation range of the abnormal sound source on the blade surface. The generation process of this distribution provides a complete geometric benchmark for cross-period spatial comparison.

[0031] Step S113: Filter the recording regions in the effective signal range of the acoustic recording intensity distribution, extract the contour boundary of the recording region along the blade span, and fit to obtain the position range.

[0032] Specifically, the effective signal range is defined based on the environmental noise baseline data collected during the steady-state operation of the unit. The lower limit of the range is set by the engineering implementation personnel in combination with the on-site turbulence intensity and sensor detection capability. Sound pressure fluctuations below the lower limit are regarded as aerodynamic noise and directly excluded. Then, the geometric outer edge of the connected region is delineated by the edge detection algorithm to extract the contour boundary. The irregular geometric outer edge is converted into a regular geometric boundary by polygon fitting operation, thereby reducing the complexity of subsequent spatial calculations.

[0033] Furthermore, establishing the location range transforms discrete sensor data into a quantifiable spatial reference, thereby providing a clear geometric reference for trajectory tracking in adjacent operating cycles.

[0034] Step S12: Based on the projected coordinates of the position range on the blade surface, pair up and associate the two position ranges with the closest spatial centroid distance in adjacent operating cycles to form a region matching pair.

[0035] Furthermore, wake disturbances cause dynamic shifts in the projected position of the defective sound source on the blade surface, and traditional fixed mesh matching is prone to multiple solution conflicts in the signal overlap area. To address this, this application uses the principle of closest spatial centroid pairing, with the geometric center of the acoustic energy distribution as a reference for cross-cycle correlation. The spatial centroid is calculated by area-weighted averaging of the projected coordinates of the position range. This pairing method adaptively tracks the sound source migration trajectory to avoid path breaks caused by forced alignment of static meshes. The generation of region matching pairs transforms discrete periodic position boundaries into spatial pairing units with clear temporal correspondences, thereby providing an accurate topological input reference for subsequent migration vector calculation.

[0036] Step S13: Calculate the coordinate difference in the spanwise direction and the projection offset in the chordwise direction of the two position ranges in the region matching pair to generate propagation continuity parameters.

[0037] Furthermore, given that recording only a single cycle position cannot reflect the migration pattern, and that the overlapping wakes cause continuous changes in the local stress state, resulting in the migration of the defect sound source propagation path along the blade length direction and lateral drift due to airflow deflection, this application summarizes the spanwise coordinate difference characterizing the path extension trend and the chordal projection offset characterizing the lateral drift amplitude to generate a propagation continuity parameter, so as to completely record the spatial migration path of the abnormal sound source in adjacent operating cycles; this parameter transforms the static position boundary into a coherent dynamic migration vector, directly reflecting the path evolution trend under wake interference, and providing a displacement constraint benchmark for subsequent continuous region correction.

[0038] Step S14: Obtain the overlap ratio and boundary fit of adjacent operating cycle regions projected onto the blade surface to form a region overlap relationship.

[0039] Specifically, the position range of the current operating cycle and the position range of the previous operating cycle are projected onto the blade surface. The ratio of the area of ​​the projection intersection to the area of ​​the position range of the previous cycle is calculated to obtain the overlap ratio. The ratio of the boundary length of the projection intersection to the boundary length of the position range of the previous cycle is extracted to obtain the boundary fit. The overlap ratio and the boundary fit together constitute the regional overlap relationship.

[0040] Furthermore, the establishment of regional overlap relationships enables the spatial interference quantification of multi-cycle propagation paths. The overlap ratio reflects the degree of overlap of energy coverage, and the boundary fit characterizes the matching accuracy of the radiation edge. The combination of the two can effectively distinguish between real defect migration and false overlap caused by instantaneous aerodynamic interference, thus providing a basis for overlap judgment to eliminate positioning jumps.

[0041] Step S15: Periodically align the propagation continuity parameter with the overlap ratio, filter out regions with consistent position change directions and continuous boundary alignment, and mark them as candidate continuous regions.

[0042] Specifically, the periodic alignment operation arranges the propagation continuous parameters and overlap ratios according to the operating cycle number to form a continuous data sequence. It traverses the continuous data sequence and checks the sign of the spanwise coordinate difference between adjacent operating cycles. Regions with consistent signs are determined to have consistent position change directions. Simultaneously, a continuous numerical sequence of boundary fit is extracted and directly input into the condition judgment process. The lower limit and upper limit of fit judgment are set as judgment references, and the lower limit and upper limit of fit judgment are determined based on the historical propagation data fluctuation range of similar units under the condition of no wake interference.

[0043] Therefore, in order to distinguish between the actual defect migration trajectory and the spurious correlation caused by aerodynamic interference under complex wake conditions, and to avoid trajectory breakage or mismatch caused by traditional single threshold judgment, this application performs hierarchical judgment and differentiated processing based on the relative relationship between the boundary fit degree value and the judgment interval, specifically making the following judgments: The first case: When the degree of boundary fit is greater than the upper limit of fit judgment, the combination of the judgment area meets the fit requirements and the combination of the area is retained to enter step S2 to perform the propagation area association processing. This judgment case corresponds to the case where the wake disturbance is weak or the propagation path is strongly constrained by the material stiffness. At this time, the boundary of the overlapping area is highly consistent and the acoustic radiation path maintains stable spatial continuity during adjacent operating cycles. The judgment retains the coherent migration trajectory that can lock the real defect and blocks the path breakage interference caused by instantaneous aerodynamic noise.

[0044] The second scenario: When the degree of boundary fit is between the lower limit and the upper limit of the fit judgment, the combination of judgment regions meets the fit requirements and the region combination is retained to enter step S2 to perform propagation region association processing. The propagation path corresponding to this judgment situation is in a critical overlapping state. At this time, the acoustic signal shows partial coupling characteristics in the overlapping area. It contains both the continuous component of the real defect migration and the boundary fluctuation component caused by the wake disturbance. Retaining the region combination can avoid missing the weak acoustic signal caused by early microcracks and ensure that the detection capability covers the boundary situation.

[0045] The third scenario: When the degree of boundary fit is less than the lower limit of fit judgment, the combination of regions does not meet the fit requirements and is directly discarded. This judgment situation corresponds to the case where the wake overlap is severe, causing the propagation path to break or shift significantly. At this time, the acoustic radiation areas of adjacent operating cycles lack effective overlap in space and the characteristic continuity is completely cut off by aerodynamic impact. Discarding the region combination can eliminate false associations caused by transient aerodynamic interference and cut off the data source of positioning jump.

[0046] Furthermore, the internal positional change relationship of the selected region combination exhibits a unidirectional progressive characteristic and the fluctuation amplitude of the positional offset relationship is controlled within the allowable deviation range. Region combinations that meet the judgment criteria are marked as candidate continuous regions to represent the acoustic propagation path that still maintains a continuous spatial migration state under wake disturbance. The output of the candidate continuous region cuts off the multi-solution conflict caused by spatial overlap, providing a data link with temporal continuity for subsequent abnormal region judgment and continuous region correction.

[0047] In some embodiments, step S2 includes steps S21 to S24: Step S21: Configure candidate continuous regions as associated nodes according to the order of the running cycles, construct periodic connection relationships, and determine the degree of boundary fit in the region overlap relationship as the connection weight between associated nodes.

[0048] Specifically, wake overlap causes acoustic propagation paths to overlap in space, making it difficult to distinguish between real defect signals and aerodynamic interference based solely on spatial location. To address this, the periodic connection relationship uses the time axis as a reference to connect discrete spatial nodes into a mesh topology, and directly maps the degree of boundary fit to the connection weight. A higher connection weight value indicates a higher degree of overlap between adjacent operating cycle paths and a stronger continuity of the sound source.

[0049] Furthermore, the process of constructing periodic connections provides a temporal framework for cross-period feature comparison. The conversion of boundary fit to connection weights realizes the quantitative mapping of spatial overlap state to association strength. The generation of these connection weights transforms the spatial fit of candidate continuous regions into an association strength index between nodes, thereby providing a basic reference for subsequent feature difference comparison and avoiding multiple solution conflicts caused by traditional fixed grid matching.

[0050] Step S22: Extract the frequency band distribution record and intensity decrease record of the local propagation acoustic text corresponding to the associated node, and determine the degree of feature difference between the connected associated node pairs in adjacent operating cycles based on the periodic connection relationship.

[0051] Specifically, a single spatial overlap index cannot eliminate artifact nodes with abrupt changes in frequency domain features. The calculation of feature difference aims to quantify the deviation of voiceprints in frequency domain energy distribution and temporal attenuation patterns between adjacent operating cycles. The synchronous extraction of frequency band distribution records and intensity decrease records constructs a multi-dimensional feature comparison dimension, breaking through the limitations of single spatial coordinate matching.

[0052] Furthermore, the calculation of the degree of feature difference is strictly limited to adjacent associated node pairs directly connected by periodic connection relationships. Its generation process cuts off the masking effect of aerodynamic noise on defect identification. The synergistic effect with the connection weight realizes the dual verification of spatial overlap and voiceprint similarity, thereby blocking false associations caused by transient interference and providing core input parameters for the construction of the regional association weight table.

[0053] In some embodiments, step S22 involves extracting the frequency band distribution record and intensity decrease record of the local propagation acoustic text corresponding to the associated node, and determining the degree of feature difference between connected pairs of associated nodes in adjacent operating cycles based on the periodic connection relationship, including steps A1 to A3: Step A1: Perform frequency band division processing on the local propagation soundprint, and count the proportion of signal records in each frequency band interval to obtain the frequency band distribution record.

[0054] Specifically, the structural acoustic signature generated by internal defects in the blade exhibits a specific energy accumulation pattern in the frequency domain, while the wake aerodynamic noise is mostly distributed in the high-frequency band. Therefore, based on the structural resonance frequency band and aerodynamic noise distribution characteristics of the blade composite material, the broadband acoustic signature signal is divided into multiple narrow band intervals. The proportion of the signal energy integral value in each narrow band interval to the total energy integral value of the entire frequency band is statistically analyzed to obtain the frequency band distribution record. This proportion characterizes the energy accumulation intensity of the defect acoustic signature in the specified frequency band.

[0055] Furthermore, the generation process of the frequency band distribution record removes broadband environmental noise interference, provides frequency domain dimension input for feature comparison of adjacent operating cycles, and its structured output cuts off the interference chain of high-frequency turbulence noise on defect identification and provides frequency domain components for the construction of feature data groups.

[0056] Step A2: Extract the peak sequence of the local propagation soundprint, calculate the ratio of the amplitude difference between adjacent peaks to the time interval, record the ratio as the amplitude decrease rate, and obtain the intensity decrease record.

[0057] Specifically, after extracting the peak sequence of locally propagating acoustic signatures along the time axis, the amplitude difference between two adjacent peaks is calculated and divided by the corresponding time interval to obtain the amplitude decrease rate, which is recorded as an intensity decrease record. This record characterizes the damping attenuation state of the acoustic signature signal inside the composite material blade and effectively distinguishes the difference in acoustic dissipation between rigid structural vibration and internal cracks in flexible materials.

[0058] Furthermore, the intensity reduction record provides a time-domain attenuation dimension input for calculating the degree of feature difference. Its establishment process excludes instantaneous high-amplitude interference caused by external airflow impact. Combined with the frequency band distribution record, it constructs a complete voiceprint feature profile to ensure the comparability of cross-cycle attenuation patterns.

[0059] Step A3: Combine the frequency band distribution records and intensity decrease records into a feature data group, and calculate the difference measure value between the feature data groups of related nodes in adjacent operating cycles as the degree of feature difference.

[0060] Specifically, the feature data set values ​​of the associated nodes in the previous cycle and the feature data set values ​​of the associated nodes in the current cycle are read. The feature data set refers to the vector set composed of frequency band distribution records and intensity decrease records arranged in a fixed dimension order. The Euclidean distance between the feature data set vectors of adjacent running cycles is calculated and directly assigned as the degree of feature difference. The smaller the Euclidean distance value, the stronger the continuity of the voiceprint feature, and the larger the value, the higher the degree of feature mutation.

[0061] Furthermore, the degree of feature difference combined with spatial connectivity weights reflects the true continuity of the sound source. Its generation process avoids misjudgment caused by single spatial overlap, provides core discrimination basis for the correlation processing of propagation area and cuts off the misleading effect of spatial artifacts on defect location.

[0062] Step S23: Match the connection weights with the degree of feature difference to generate a region association weight table.

[0063] Specifically, high spatial overlap does not necessarily mean true continuity of the sound source. Aerodynamic interference can also generate highly overlapping regions. Therefore, the matching process will jointly calculate the inverse mapping value of the connection weight value and the feature difference degree. The lower the feature difference degree value, the stronger the voiceprint continuity. The inverse mapping value is obtained by nonlinearly transforming the feature difference degree using a preset mapping function with an inverse proportional transformation rule or a negative exponential decay rule, so as to strengthen the weight ratio of low difference nodes.

[0064] Furthermore, collaborative computing performs a product operation on the connection weights and the reverse mapping values ​​to generate cell values ​​in the regional association weight table. These values ​​simultaneously represent the joint strength of spatial path overlap and voiceprint feature continuity, achieving dual-dimensional fusion to overcome the limitations of single-dimensional judgment, thereby providing a comprehensive evaluation index and direct data source for subsequent temporal path screening.

[0065] Step S24: Based on the periodic connection relationship, the associated nodes connected in adjacent running cycles are sequentially connected to generate node connection paths; the connection weights corresponding to the node connection paths are read according to the regional association weight table, the connection weights are calculated, and the node path scores are obtained; the node path scores are sorted, the node connection path with the highest node path score is selected as the propagation association sequence, and the extreme values ​​of the boundary fitting degree corresponding to the associated nodes in the node connection path are extracted as regional overlap parameters.

[0066] Specifically, the node connection paths formed by connecting adjacent running cycle related nodes according to the periodic connection relationship are traversed. The connection weights corresponding to each path are read from the regional association weight table and accumulated to obtain the node path score. After sorting all node path scores in descending order, the node connection path with the highest score is selected as the propagation association sequence. The maximum and minimum values ​​of the boundary fitting degree values ​​corresponding to each related node in the path are extracted as the boundary fitting degree extreme values ​​and directly assigned as the regional overlap parameter.

[0067] Furthermore, transient aerodynamic interference nodes are accompanied by abrupt changes in acoustic signature and lack a continuous evolution pattern in spatial location. The corresponding cumulative weight value in the regional association weight table shows a discrete fluctuation state and is at the end of the sorting. The real defect sound source maintains a stable spatial migration trajectory and frequency domain attenuation state in the wake overlap area. Its cumulative weight value maintains a continuous high-level accumulation along the time axis. The numerical descending sorting and first-place selection action directly lock the high-cumulative link and retain the real propagation trajectory with physical continuity.

[0068] Furthermore, the output of the propagation association sequence cuts off the multiple solution conflicts caused by spatial overlap, providing a data link with temporal continuity for the determination of abnormal propagation areas. Its synchronous generation with the regional overlap parameters locks the temporal continuity and spatial overlap, completely eliminating the positioning jump defect under the wake disturbance.

[0069] Better than traditional wind turbine blade monitoring methods that rely on fixed spatial grids for cross-cycle feature matching, fixed grid matching incorrectly binds similar acoustic signatures from different operating cycles in the wake overlap area, causing jumps in the positioning results. By constructing periodic connection relationships, the structure of discrete nodes is transformed into temporal topology. The mapping of boundary fit to connection weights completes the quantitative calibration of spatial overlap. The extraction of frequency band distribution records and intensity decrease records removes broadband noise interference. The calculation of feature differences enables accurate measurement of acoustic signature continuity between adjacent operating cycles. The collaborative calculation of connection weights and reverse mapping values ​​generates a regional association weight table. The cumulative calculation and numerical sorting of node path scores lock the true defect propagation trajectory. The synchronous output of propagation association sequences and regional overlap parameters provides dual data support with temporal continuity and spatial overlap for the determination of abnormal propagation areas. The processing flow completely cuts off the data source of positioning jumps from the feature association layer.

[0070] S3: Determine the abnormal propagation region based on the propagation association sequence and region overlap parameters; In some embodiments, step S3 includes steps S31 to S34: Step S31: Perform periodic signal recording superposition processing on each associated node in the propagation association sequence to obtain the cumulative value of the voiceprint feature corresponding to each associated node.

[0071] Specifically, the acoustic signature signal of a single operating cycle is susceptible to interference from instantaneous airflow fluctuations, resulting in random fluctuations. To address this, the periodic signal recording superposition processing numerically accumulates the acoustic signature records of the same associated node within consecutive operating cycles, thereby reducing random noise components and highlighting the stable radiation of defective sound source signals.

[0072] Furthermore, the process of generating cumulative voiceprint features enables the temporal aggregation of multi-cycle signal energy. Its establishment process reduces the probability of instantaneous aerodynamic noise interfering with defect identification and provides a quantitative benchmark for subsequent correction of overlapping interference, thereby cutting off the interference chain of misjudgment of single-cycle signals.

[0073] Step S32: Generate the overlap interference correction coefficient based on the extreme value of the boundary fitting degree in the region overlap parameters, and perform numerical conversion processing on the cumulative value of voiceprint features to obtain the converted feature cumulative value.

[0074] In some embodiments, step S32 includes steps B1 to B5: Step B1: From the associated nodes contained in the propagation associated sequence, filter out the associated nodes that have overlapping blade surface projections within the same running cycle and determine them as candidate nodes.

[0075] Specifically, the wake overlap region exhibits the physical phenomenon of multiple propagation paths simultaneously radiating acoustic signatures, leading to overlapping spatial projections of these paths, and the sensor-received signals contain multi-source mixed components. To address this, this application identifies associated nodes with overlapping projections within the same operating cycle from the propagation correlation sequence, clarifies the spatial interference range, and obtains the regional overlap parameters corresponding to each candidate node.

[0076] Furthermore, the identification operation of candidate nodes delineates the boundary of overlapping interference. The acquisition process provides a spatial overlap reference for subsequent numerical conversion processing and avoids erroneous merging of voiceprint signals in the intersection area.

[0077] Step B2: Extract the extreme values ​​of the boundary fitting degree of the region overlap parameters corresponding to each candidate node, and use the ratio of the extreme value of the boundary fitting degree to the benchmark fitting value as the overlap interference correction coefficient.

[0078] Specifically, the maximum and minimum values ​​of the boundary fitting degree recorded by each candidate node are read as the extreme values ​​of the boundary fitting degree, and the ratios of these values ​​with the baseline fitting value are calculated. The larger the ratio value, the more obvious the boundary inconsistency and multi-source interference phenomenon exist in the overlapping area. The overlap interference correction coefficient is directly generated based on this ratio, so that the larger the ratio value, the higher the correction coefficient, and the smaller the ratio value, the lower the correction coefficient.

[0079] Furthermore, the extraction of the extreme values ​​of boundary overlap enables the conversion of spatial overlap into numerical correction weights, laying a data foundation for multi-source signal separation.

[0080] Step B3: Evaluate the continuous state of the characteristic path of the candidate node and the associated node in the adjacent running cycle, and determine the node whose continuous state of the characteristic path of the candidate node undergoes a step change as a transient interference node.

[0081] Specifically, the position coordinates and voiceprint feature parameters of the candidate node and the adjacent running cycle associated nodes are read, and the deviation vector between the candidate node coordinates and the adjacent node coordinates is calculated. When the magnitude of the deviation vector exceeds the set continuous tolerance range and the voiceprint feature parameters change abruptly, it is determined that the continuous state of the feature path has undergone a step change, and the corresponding node is directly marked as a transient interference node.

[0082] Furthermore, the judgment logic for transient interference nodes removes the instantaneous path breakage caused by wake turbulence. Its identification process cuts off the interference of discontinuous acoustic signals on the cumulative value and eliminates the misleading effect of spatial artifacts on defect location.

[0083] Step B4: Eliminate the cumulative voiceprint feature values ​​corresponding to transient interference nodes, and select the nodes whose feature path continuously evolves smoothly from the remaining candidate nodes as the main propagation nodes.

[0084] Specifically, the position coordinate sequence and voiceprint feature parameter sequence of the remaining candidate nodes are read, and the coordinate offset and feature difference value between adjacent nodes are calculated. When the coordinate offset is within the set stable evolution range and the feature difference value is lower than the set feature tolerance range, it is determined that the feature path continuity state maintains stable evolution, and the corresponding node is marked as the main propagation node.

[0085] Furthermore, the screening process of the main propagation node identifies real defect sound sources with physical continuity, whose stable evolution state corresponds to the acoustic radiation pattern caused by the stable expansion of the defect structure, thus providing pure data input for subsequent accurate conversion.

[0086] Step B5: Perform numerical conversion processing on the cumulative value of the acoustic feature corresponding to the main propagation node based on the overlap interference correction coefficient to obtain the converted feature cumulative value.

[0087] Specifically, the cumulative value of the voiceprint feature and the overlap interference correction coefficient corresponding to the main propagation node are read. The cumulative value of the voiceprint feature is divided by the overlap interference correction coefficient to perform a numerical conversion operation. The higher the overlap interference correction coefficient, the greater the reduction, and the lower the coefficient, the smaller the reduction. The result after the conversion operation is recorded as the converted feature cumulative value.

[0088] Furthermore, the conversion calculation eliminates the energy inflation effect caused by spatial overlap and restores the true radiation intensity of a single defective sound source, providing high-confidence data support for the determination of abnormal propagation areas.

[0089] Step S33: Compare the calculated cumulative value of the feature with the historical baseline sequence, and filter out the associated nodes whose calculated values ​​are higher than the upper limit of the distribution of the historical baseline sequence within the continuous running period to obtain the set of abnormal associated nodes.

[0090] Specifically, the historical baseline sequence is a pre-stored benchmark dataset for the defect-free operation period, constructed from the statistical distribution range of the cumulative voiceprint feature values ​​of each spatial node under defect-free conditions. The time-series comparison operation compares the converted feature cumulative values ​​with the upper limit of the historical baseline sequence distribution. When the comparison result continuously exceeds the upper limit and spans multiple consecutive operating cycles, it is determined that the converted value is in a continuously high state.

[0091] Furthermore, the extraction process of the abnormal associated node set cuts off the interference of steady-state background noise on the anomaly determination. Its generation realizes the quantitative separation of abnormal signals from background noise and provides a range of candidate targets for subsequent spatial clustering feature evaluation.

[0092] Step S34: Calculate the coordinate difference of adjacent associated nodes in the blade spanwise direction in the abnormal associated node set, divide the associated nodes whose coordinate difference is less than the clustering distance threshold into spatial clustering node groups, and determine the blade spanwise coordinate interval covered by the spatial clustering node group as the abnormal propagation area.

[0093] Specifically, the clustering spacing threshold is calibrated based on the spanwise propagation length of typical cracks inside the blade and the spatial resolution of the acoustic sensor. The spanwise coordinate position of each node in the abnormal associated node set is read and the coordinate difference between adjacent nodes is calculated. When the difference is less than the clustering spacing threshold, the corresponding adjacent nodes are divided into spatial clustering node groups. All node groups are traversed and their spanwise coordinate coverage area is extracted and directly marked as the abnormal propagation area.

[0094] Furthermore, the spatial aggregation characteristic evaluation process excludes randomly distributed environmental noise nodes, whose states correspond to the physical laws of the stable existence of a single defect source inside the composite material, completing the accurate mapping from time-series signals to spatial locations and finally locking the geometric boundaries of the defect source.

[0095] In contrast, traditional monitoring methods rely on fixed spatial grid matching under wake overlap conditions, which can easily lead to incorrect binding of similar acoustic signatures from different operating cycles in the overlapping area, causing jumps in the positioning results. In this stage, multi-cycle energy aggregation is achieved through periodic signal recording and superposition processing to reduce instantaneous aerodynamic noise interference. The overlap interference correction coefficient and numerical conversion processing remove the energy inflation caused by the intersection and overlap of multiple source paths. The continuous state assessment of characteristic paths cuts off the false correlation caused by step jumps. The temporal comparison of historical baseline sequences and spatial aggregation feature assessment double filter random noise. The processing flow stably outputs the true location of the defect under wake disturbance and path overlap conditions, completely eliminates the positioning jump defect from the data judgment layer, and provides spatially deterministic abnormal propagation area input for continuous area correction.

[0096] S4: Perform continuous region correction on the abnormal propagation region based on the propagation continuity parameter and the region overlap parameter, and output the abnormal propagation result data.

[0097] In some embodiments, step S4 includes steps S41 to S45: Step S41: Extract the position change relationship sequence from the propagation continuous parameters, determine the expected displacement vector of adjacent operating cycles based on the position change relationship sequence, and establish a displacement evolution benchmark; Specifically, wake overlap and material stress alternation cause periodic drift in the sound source location coordinates. If the original position sequence is directly transmitted downstream, it will cause coordinate abrupt change interference. To address this, this application extracts the position change relationship sequence from the propagation continuity parameters and derives the expected displacement vector of adjacent operating cycles to establish a displacement evolution benchmark. This benchmark is used to constrain the extension direction and amplitude threshold of subsequent coordinate corrections.

[0098] Furthermore, the establishment of the displacement evolution benchmark incorporates discrete jump coordinates into the calculation scope of continuous evolution trajectory, realizes the active constraint of physical laws on positioning drift, and provides a stable calculation reference for subsequent position adjustment processing.

[0099] Step S42: Combining the displacement evolution benchmark, the sound velocity parameters of the blade composite material, and the curvature of the local surface, perform position adjustment processing on the original position sequence of the abnormal propagation region and output the preliminary adjustment sequence; Specifically, the sound velocity parameters of the blade composite material characterize the propagation rate of sound waves in the alternating stress field, and the curvature of the local surface defines the propagation path of sound waves along the geometric surface of the blade. The position adjustment process takes the displacement evolution benchmark, the material sound velocity parameters and the curvature of the local surface as the calculation input, and performs coordinate smoothing operation on the original position sequence to output the preliminary adjustment sequence.

[0100] Furthermore, this processing breaks through the spatial limitations of fixed grid matching, integrating macroscopic acoustic characteristics with microscopic geometric morphology, and providing a smooth benchmark with working condition adaptability for subsequent jump point identification.

[0101] In some embodiments, step S42 combines the displacement evolution benchmark, the sound velocity parameters of the blade composite material, and the curvature of the local surface to perform position adjustment processing on the original position sequence of the abnormal propagation region, and outputs a preliminary adjustment sequence, including steps S421 to S423: Step S421: Obtain the wake turbulence intensity value and blade root bending moment load data from the unit operation record; Specifically, the operating status of the wind turbine directly determines the stress environment of the blades and the intensity of airflow disturbance. The numerical quantification of the wake turbulence intensity results in the impact level of the front wake turbulence borne by the turbine unit. The bending moment load data at the blade root reflects the amplitude of the alternating aerodynamic bending moment borne by the blade as a whole. The data acquisition operation provides a real-time input source for stress inversion and operating condition mapping.

[0102] Furthermore, the retrieval of operation record data establishes a connection between the unit's macroscopic operating conditions and microscopic acoustic positioning, ensuring that position adjustment processing has a basis for dynamic operating condition adaptation and avoiding environmental mismatch defects caused by static parameter correction.

[0103] Step S422: Determine the stress distribution along the span of the blade based on the bending moment load data at the blade root, and determine the sound velocity adjustment value corresponding to the stress distribution according to the material sound velocity correspondence. Specifically, the bending moment load data at the blade root is read and substituted into the structural mechanics transfer model to calculate the distribution of normal stress and shear stress on each section along the blade span. Then, the material sound velocity correspondence table is read and the stress distribution values ​​are mapped to the table to obtain the sound velocity change ratio, which is recorded as the sound velocity adjustment value.

[0104] Furthermore, the calculation of stress distribution enables cross-scale conversion of macroscopic loads to microscopic acoustic parameters, and the generation of sound velocity adjustment values ​​accurately reflects the acoustoelastic effect of composite materials under alternating stress, eliminating the propagation path calculation deviation caused by stress redistribution.

[0105] Step S423: Adjust the execution parameters of the position adjustment process according to the wake turbulence intensity value. The position adjustment process generates a preliminary adjustment sequence based on the adjusted execution parameters under the condition of wake overlap. Specifically, a continuous mapping function between the wake turbulence intensity value and the parameter adjustment range is established to convert the turbulence intensity into a step size tolerance amplification factor and a data smoothing weight attenuation factor. When the turbulence intensity increases, the step size tolerance is gradually increased while the data smoothing weight is reduced simultaneously. Conversely, the step size tolerance is reduced while the weight is increased. The displacement evolution benchmark, sound velocity adjustment value, and adjusted execution parameters are input into the iterative calculation module to generate a preliminary adjustment sequence under the wake coincidence condition.

[0106] Furthermore, continuous adjustment of the execution parameters enables precise matching between the filtering intensity and the airflow disturbance level, avoiding excessive smoothing under strong turbulence and insufficient correction under weak turbulence, thus ensuring the numerical stability of the coordinate correction in the overlapping area.

[0107] Step S43: Identify position jump points by using the extreme values ​​of boundary fit in the region overlap parameters, and read the stable coordinate data of adjacent running cycles before and after the position jump points; Specifically, overlapping propagation paths in adjacent operating cycles can easily cause coordinate abrupt changes in the positioning algorithm at the overlapping boundary, thereby disrupting the continuity of the defect evolution trajectory. This application extracts the extreme value of the boundary fit and calculates the sum of the absolute deviations of the differences between adjacent cycles as the fluctuation amplitude. When the amplitude exceeds the set fluctuation threshold, the corresponding cycle node is marked as the position jump point, and the stable coordinate data of the adjacent operating cycles before and after the point are read.

[0108] Furthermore, the process of identifying position jump points accurately locates the coordinate break points caused by overlapping interference, providing a clear target segment for subsequent interpolation.

[0109] Step S44: Generate continuous transition coordinates based on stationary coordinate data, replace the jump coordinates in the initial adjustment sequence with continuous transition coordinates, arrange the coordinate data in the order of the running cycle, and generate a position correction sequence; Specifically, based on the read stable coordinate data, an interpolation algorithm is used to generate continuous transition coordinates. The jump coordinates in the initial adjustment sequence are directly replaced with these continuous transition coordinates, and the coordinate data are rearranged according to the order of the running cycle to generate a position correction sequence.

[0110] Furthermore, the generation of continuous transition coordinates replaces jump coordinates with mathematical interpolation and logical smoothing, restoring the geometric continuity of the defect sound source migration trajectory and completely severing the positioning jump chain at the overlapping boundary.

[0111] Step S45: Project the position correction sequence onto the blade's three-dimensional spatial coordinate system and output anomaly propagation result data containing the corrected spanwise position coordinates, anomaly extension direction, and regional confidence weights.

[0112] Specifically, the spatial coordinate data in the position correction sequence is projected onto the surface mesh node corresponding to the three-dimensional spatial coordinate system of the blade. The spanwise coordinate values ​​of the projection points are extracted as the corrected spanwise position coordinates. The direction vector of the line connecting adjacent projection points is calculated as the direction of anomaly extension. The weighted product of coordinate smoothness index and cumulative intensity of acoustic features is calculated as the regional confidence weight. Finally, the anomaly propagation result data is integrated and output.

[0113] Furthermore, the 3D projection process transforms temporal coordinates into quantifiable spatial defect distributions. The synchronized output of corrected coordinates, extension directions, and confidence weights clearly defines the geometric boundaries, expansion trends, and confidence levels of defects, providing a direct basis for operation and maintenance scheduling decisions.

[0114] Ideally, traditional blade monitoring systems rely on fixed filtering parameters under overlapping wake conditions, which cannot adapt to the real-time changes in dynamic stress fields and turbulence intensity, leading to frequent jumps in positioning results. This stage establishes a displacement evolution benchmark by extracting the position change relationship sequence, and dynamically adjusts the execution parameters in combination with the operating conditions to achieve precise coupling between filtering intensity and disturbance level. It uses the extreme value of boundary fit to accurately remove coordinate breaks caused by overlapping interference, and the generation of continuous transition coordinates restores the geometric continuity of the migration trajectory. The output of abnormal propagation result data enables three-dimensional synchronous delivery of defect location, expansion trend and confidence level. The processing flow completely eliminates the positioning jump defects under overlapping wake conditions from the coordinate correction layer.

[0115] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0116] Based on the same inventive concept, this application also provides an online monitoring system for wind turbine blades. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the online monitoring system for wind turbine blades provided below can be found in the limitations of the online monitoring method for wind turbine blades described above, and will not be repeated here.

[0117] Reference Figure 2 The online monitoring method for wind turbine blades in this application uses an array of acoustic fingerprint sensors deployed along the blade spanwise to collect time-series sound pressure records in real time during each operating cycle. An embedded acquisition module performs time-segmentation of the time-series sound pressure records in sync with the blade rotation phase to determine the reference sound pressure record value for each measuring point. This reference sound pressure record value is then transmitted wirelessly to an edge computing node or a centralized control server. The centralized control server generates an acoustic fingerprint intensity distribution by spatially mapping the reference sound pressure record value to the blade surface measuring point layout diagram. It extracts the location range and calculates the position change relationship and position offset relationship between adjacent operating cycles. The location range is compared for overlap to obtain the overlap ratio and boundary fit. The position change relationship, position offset relationship, and overlap ratio are periodically aligned to filter out candidate continuous regions.

[0118] Specifically, the centralized control server constructs periodic connectivity relationships based on candidate continuous regions and determines the degree of feature differences. After generating a regional association weight table, it filters propagation association sequences and generates overlapping interference correction coefficients based on the extreme values ​​of boundary fit to eliminate transient interference nodes. Simultaneously, it dynamically adjusts the execution parameters of position adjustment processing by combining wake turbulence intensity values ​​retrieved from the data storage system with blade root bending moment load data. It outputs a position correction sequence and maps it to the blade's three-dimensional structure diagram, ultimately generating anomaly propagation result data that includes corrected spanwise position coordinates, anomaly propagation direction, and regional confidence weights. The intelligent inspection terminal can access the centralized control server or cloud storage via the network to obtain anomaly propagation result data, providing maintenance personnel with visualized monitoring and decision support for blade defects.

[0119] In one exemplary embodiment, such as Figure 3 As shown, an online monitoring system for wind turbine blades is provided, including: The trajectory construction module calculates the continuous change trajectory of the propagation area and determines candidate continuous areas within adjacent operating cycles during the wind turbine blade operation acquisition process, based on the distribution status of the signal propagation area and the regional overlap relationship between the signal propagation areas, and obtains propagation data. The propagation data includes propagation continuity parameters, regional overlap relationship and candidate continuous areas. The association processing module performs propagation region association processing on the local propagation voiceprints corresponding to the running cycle based on candidate continuous regions and region overlap relationships to obtain association data, which includes propagation association sequences and region overlap parameters. The anomaly detection module determines the anomaly propagation area based on the propagation association sequence and region overlap parameters. The continuous correction module performs continuous region correction on the abnormal propagation region based on the propagation continuity parameter and the region overlap parameter, and outputs the abnormal propagation result data.

[0120] The modules in the aforementioned online monitoring system for wind turbine blades can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0121] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements an online monitoring method for wind turbine blades. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0122] Those skilled in the art will understand that Figure 4 The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer devices on which the present application is applied. Specific computer devices may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0123] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0124] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0125] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0126] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0127] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0128] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0129] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for online monitoring of wind turbine blades, characterized in that, include: During adjacent operating cycles of the wind turbine blade operation data acquisition process, the continuous change trajectory of the propagation area is calculated and candidate continuous areas are determined based on the distribution status of the signal propagation area and the regional overlap relationship between the signal propagation areas, thus obtaining propagation data. The propagation data includes propagation continuity parameters, regional overlap relationship and candidate continuous areas. Based on candidate continuous regions and region overlap relationships, propagation region association processing is performed on the local propagation voiceprints corresponding to the running cycle to obtain association data, which includes propagation association sequence and region overlap parameters. The abnormal propagation region is determined based on the propagation association sequence and the region overlap parameter; Based on the propagation continuity parameter and the region overlap parameter, continuous region correction is performed on the abnormal propagation region, and abnormal propagation result data is output.

2. The online monitoring method for wind turbine blades according to claim 1, characterized in that, During the wind turbine blade operation data acquisition process, within adjacent operating cycles, based on the signal propagation area distribution and the overlap relationship between signal propagation areas, the continuous change trajectory of the propagation area is calculated and candidate continuous areas are determined to obtain propagation data, including: Extract the position range of each signal propagation region along the blade span within adjacent operating cycles to form the signal propagation region distribution state; Based on the projected coordinates of the position range on the blade surface, the two position ranges with the closest spatial centroid distance in adjacent operating cycles are paired and associated to form a region matching pair; Calculate the coordinate difference in the spanwise direction and the projection offset in the chordwise direction of the two position ranges in the region matching pair to generate the propagation continuity parameter; Obtain the overlap ratio and boundary fit of adjacent operating cycle regions projected onto the blade surface to form the region overlap relationship; The propagation continuity parameter and the overlap ratio are periodically aligned, and regions with consistent position change directions and continuous boundary adhesion are selected and marked as candidate continuous regions.

3. The online monitoring method for wind turbine blades according to claim 2, characterized in that, The step of extracting the position range of each signal propagation region along the blade spanwise within adjacent operating cycles to form the signal propagation region distribution status includes: The temporal sound pressure records output by the acoustic fingerprint sensor array deployed along the blade span are obtained in each operating cycle. The data segments of each operating cycle are extracted with the blade rotation phase as the alignment reference, and the reference sound pressure record value of each acoustic fingerprint sensor measurement point is determined. The reference sound pressure recording value is assigned to the three-dimensional coordinate position corresponding to the blade surface measurement point layout diagram, and the acoustic pattern recording intensity distribution along the blade span is obtained by spatial interpolation. The recording regions within the effective signal range of the acoustic signature intensity distribution are selected, and the contour boundaries of the recording regions along the blade span are extracted and fitted to obtain the position range.

4. The online monitoring method for wind turbine blades according to claim 1, characterized in that, The process of performing propagation region association processing on the local propagation voiceprints corresponding to the operating cycle based on candidate continuous regions and region overlap relationships yields associated data, including: The candidate continuous regions are configured as associated nodes according to the order of the operation cycle, a periodic connection relationship is constructed, and the degree of boundary fit in the region overlap relationship is determined as the connection weight between associated nodes. Extract the frequency band distribution record and intensity decrease record of the local propagation acoustic text corresponding to the associated node, and determine the degree of feature difference between the connected associated node pairs in adjacent operating cycles based on the periodic connection relationship; The connection weights are matched with the degree of feature difference to generate a region association weight table; Based on the aforementioned periodic connection relationship, the associated nodes connected in adjacent operating cycles are sequentially linked together to generate a node connection path; The connection weights corresponding to the node connection paths are read from the region association weight table, and the connection weights are calculated to obtain the node path score. The node path scores are sorted, and the node connection path with the highest score is selected as the propagation association sequence. The extreme values ​​of the boundary fit of the associated nodes in the node connection path are extracted as the region overlap parameters.

5. The online monitoring method for wind turbine blades according to claim 4, characterized in that, The step of extracting the frequency band distribution record and intensity decrease record of the local propagation acoustic text corresponding to the associated node, and determining the degree of feature difference between connected pairs of associated nodes in adjacent operating cycles based on the periodic connection relationship, includes: The local propagation acoustic text is subjected to frequency band division processing, and the signal recording ratio in each frequency band interval is statistically analyzed to obtain the frequency band distribution record. Extract the peak sequence of the local propagation soundprint, calculate the ratio of the amplitude difference between adjacent peaks to the time interval, record the ratio as the amplitude decrease rate, and obtain the intensity decrease record; The frequency band distribution record and the intensity decrease record are combined into a feature data group, and the difference measure value between the feature data groups of related nodes in adjacent operating cycles is calculated as the feature difference degree.

6. The online monitoring method for wind turbine blades according to claim 1, characterized in that, The step of determining the abnormal propagation region based on the propagation association sequence and the region overlap parameter includes: Perform periodic signal recording superposition processing on each associated node in the propagation association sequence to obtain the cumulative value of the voiceprint feature corresponding to each associated node; Based on the extreme values ​​of boundary fitting in the region overlap parameters, an overlap interference correction coefficient is generated, and numerical conversion processing is performed on the cumulative value of voiceprint features to obtain the converted cumulative value of features. The calculated cumulative value of the feature is compared with the historical baseline sequence, and the associated nodes whose calculated values ​​are higher than the upper limit of the distribution of the historical baseline sequence within a continuous running period are selected to obtain a set of abnormal associated nodes; Calculate the coordinate difference in the blade span of adjacent associated nodes in the set of abnormal associated nodes, divide the associated nodes whose coordinate difference is less than the clustering distance threshold into spatial clustering node groups, and determine the blade span of coordinate interval covered by the spatial clustering node groups as the abnormal propagation area.

7. The online monitoring method for wind turbine blades according to claim 6, characterized in that, The process of generating overlap interference correction coefficients based on the extreme values ​​of boundary fit in the region overlap parameters, and performing numerical conversion processing on the cumulative value of voiceprint features to obtain the converted cumulative value of features includes: From the associated nodes contained in the propagation association sequence, the associated nodes that have overlapping blade surface projections within the same operating cycle are selected and determined as candidate nodes; Extract the extreme values ​​of the boundary fitting degree of the region overlap parameters corresponding to each candidate node, and use the ratio of the extreme values ​​of the boundary fitting degree to the reference fitting value as the overlap interference correction coefficient. Evaluate the continuous state of the characteristic path of the candidate node and the associated node in the adjacent running cycle, and determine the node in the candidate node whose continuous state of the characteristic path undergoes a step change as a transient interference node. The cumulative value of the acoustic signature feature corresponding to the transient interference node is removed, and the node whose feature path continues to evolve smoothly among the remaining candidate nodes is selected as the main propagation node. Based on the overlapping interference correction coefficient, the cumulative value of the voiceprint feature corresponding to the main propagation node is numerically converted to obtain the converted feature cumulative value.

8. The online monitoring method for wind turbine blades according to claim 1, characterized in that, The step of performing continuous region correction on the abnormal propagation region based on the propagation continuity parameter and the region overlap parameter, and outputting abnormal propagation result data, includes: The position change relationship sequence is extracted from the propagation continuity parameters, and the expected displacement vector of adjacent operating cycles is determined based on the position change relationship sequence to establish a displacement evolution benchmark. Combining the displacement evolution benchmark, the sound velocity parameters of the blade composite material, and the curvature of the local surface, the original position sequence of the abnormal propagation region is subjected to position adjustment processing, and a preliminary adjustment sequence is output. The extreme value of the boundary fit in the region overlap parameter is used to identify the position jump point, and the stable coordinate data of the adjacent running cycles before and after the position jump point are read. Based on the stable coordinate data, continuous transition coordinates are generated. The jump coordinates in the preliminary adjustment sequence are replaced with the continuous transition coordinates. The coordinate data are arranged in the order of the running cycle to generate a position correction sequence. The position correction sequence is projected onto the blade's three-dimensional spatial coordinate system, and the anomaly propagation result data, which includes the corrected spanwise position coordinates, the anomaly extension direction, and the region confidence weight, is output.

9. The online monitoring method for wind turbine blades according to claim 8, characterized in that, The process combines the displacement evolution benchmark, the sound velocity parameters of the blade composite material, and the curvature of the local surface to perform position adjustment processing on the original position sequence of the abnormal propagation region, outputting a preliminary adjustment sequence, including: Obtain wake turbulence intensity values ​​and blade root bending moment load data from the unit's operation records; Based on the bending moment load data at the blade root, the stress distribution along the blade span is determined, and the sound velocity adjustment value corresponding to the stress distribution is determined according to the material sound velocity correspondence. The execution parameters of the position adjustment process are adjusted according to the wake turbulence intensity value, and the position adjustment process generates the preliminary adjustment sequence based on the adjusted execution parameters under the condition of wake overlap.

10. A wind turbine blade online monitoring system, employing the wind turbine blade online monitoring method as described in any one of claims 1 to 9, characterized in that, include: The trajectory construction module calculates the continuous change trajectory of the propagation area and determines candidate continuous areas based on the distribution status of the signal propagation area and the regional overlap relationship between the signal propagation areas during adjacent operating cycles in the wind turbine blade operation acquisition process, thereby obtaining propagation data. The propagation data includes propagation continuity parameters, regional overlap relationship and candidate continuous areas. The association processing module performs propagation region association processing on the local propagation voiceprints corresponding to the running cycle based on candidate continuous regions and region overlap relationships to obtain association data, which includes propagation association sequences and region overlap parameters. The anomaly detection module determines the abnormal propagation region based on the propagation association sequence and the region overlap parameter; The continuous correction module performs continuous region correction on the abnormal propagation region based on the propagation continuity parameter and the region overlap parameter, and outputs abnormal propagation result data.

Images